deepskies / DeepCMBsim

MIT License
2 stars 0 forks source link

Sketch workflow #2

Closed cavestruz closed 1 year ago

cavestruz commented 2 years ago

Add sketch to workflow in the readme

cavestruz commented 2 years ago

suggest to do this in a separate branch, e.g. issue/2/sketched-workflow

cavestruz commented 2 years ago

image

cavestruz commented 2 years ago

config_parser.py .... cosmo_power_spectra.py ....def call_camb(cosmo) ....def fit_spectra(spectrum)

generate_images.py ...def generatett(spectrum): ...def generate....

reconstruction_methods.py ...def fft_calc(image): ...def.... reconstruction_method = {'fft': fft_calc}

cavestruz commented 2 years ago

The below should live in a example_config directory

config.ini (or yml, whatever) [cosmo] omegam : .... r : ....

[image] num_pix : 1024 restrict_image_types : ["tt", ""] # None reconstruction_method : fft

cavestruz commented 2 years ago

python run_cmb_sim.py configdir ..... reads in config file in configdir .... goes through steps so there are power spectra in a file (or folder) in configdir, and images in a file (or folder) in configdir

cavestruz commented 2 years ago

To get sims for SBI approach, or some other ML approach: --> configfile to contain a list of cosmological parameters (really just vary over r or Alens), if both, have a list of cartesian product --> once run, we get spectra for each item in the cartesian product and images for each cartesian product item

.... SBI takes in the list of params (e.g. cartesian product) in parallel with either the spectra or 2d images corresponding to each item, some are used to train, some to test and see how close the "measured" r or Alens might be.

samueldmcdermott commented 1 year ago

subsumed by #47 and by @humnaawan 's work here